[PATCH v3 0/3] ext4: fix some bugs in online resize

[PATCH v3 0/3] ext4: fix some bugs in online resize

[Baokun Li]

V1->V2:
    Replace s_first_data_block with ext4_group_first_block_no() in patch 3
    to avoid type warning.(Reported-by: kernel test robot <lkp@intel.com>)
V2->V3:
    Modify the patch2 scheme and retain s_block_group_nr.

Baokun Li (3):
  ext4: fix bad checksum after online resize
  ext4: fix corrupt backup group descriptors after online resize
  ext4: fix corruption when online resizing a 1K bigalloc fs

 fs/ext4/resize.c | 32 ++++++++++++++++++++------------
 1 file changed, 20 insertions(+), 12 deletions(-)

-- 
2.31.1

[PATCH v3 1/3] ext4: fix bad checksum after online resize

[Baokun Li]

When online resizing is performed twice consecutively, the error message
"Superblock checksum does not match superblock" is displayed for the
second time. Here's the reproducer:

	mkfs.ext4 -F /dev/sdb 100M
	mount /dev/sdb /tmp/test
	resize2fs /dev/sdb 5G
	resize2fs /dev/sdb 6G

To solve this issue, we moved the update of the checksum after the
es->s_overhead_clusters is updated.

Fixes: 026d0d27c488 ("ext4: reduce computation of overhead during resize")
Fixes: de394a86658f ("ext4: update s_overhead_clusters in the superblock during an on-line resize")
Signed-off-by: Baokun Li <libaokun1@huawei.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
---
 fs/ext4/resize.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/fs/ext4/resize.c b/fs/ext4/resize.c
index 46b87ffeb304..cb99b410c9fa 100644
--- a/fs/ext4/resize.c
+++ b/fs/ext4/resize.c
@@ -1476,8 +1476,6 @@ static void ext4_update_super(struct super_block *sb,
 	 * active. */
 	ext4_r_blocks_count_set(es, ext4_r_blocks_count(es) +
 				reserved_blocks);
-	ext4_superblock_csum_set(sb);
-	unlock_buffer(sbi->s_sbh);
 
 	/* Update the free space counts */
 	percpu_counter_add(&sbi->s_freeclusters_counter,
@@ -1513,6 +1511,8 @@ static void ext4_update_super(struct super_block *sb,
 		ext4_calculate_overhead(sb);
 	es->s_overhead_clusters = cpu_to_le32(sbi->s_overhead);
 
+	ext4_superblock_csum_set(sb);
+	unlock_buffer(sbi->s_sbh);
 	if (test_opt(sb, DEBUG))
 		printk(KERN_DEBUG "EXT4-fs: added group %u:"
 		       "%llu blocks(%llu free %llu reserved)\n", flex_gd->count,
-- 
2.31.1

[PATCH v3 2/3] ext4: fix corrupt backup group descriptors after online resize

[Baokun Li]

In commit 9a8c5b0d0615 ("ext4: update the backup superblock's at the end
of the online resize"), it is assumed that update_backups() only updates
backup superblocks, so each b_data is treated as a backupsuper block to
update its s_block_group_nr and s_checksum. However, update_backups()
also updates the backup group descriptors, which causes the backup group
descriptors to be corrupted.

The above commit fixes the problem of invalid checksum of the backup
superblock. The root cause of this problem is that the checksum of
ext4_update_super() is not set correctly. This problem has been fixed
in the previous patch ("ext4: fix bad checksum after online resize").

However, we do need to set block_group_nr for the backup superblock in
update_backups(). When a block is in a group that contains a backup
superblock, and the block is the first block in the group, the block is
definitely a superblock. We add a helper function that includes setting
s_block_group_nr and updating checksum, and then call it only when the
above conditions are met to prevent the backup group descriptors from
being incorrectly modified.

Fixes: 9a8c5b0d0615 ("ext4: update the backup superblock's at the end of the online resize")
Signed-off-by: Baokun Li <libaokun1@huawei.com>
---
V2->V3:
    Modify the scheme and retain s_block_group_nr.

 fs/ext4/resize.c | 22 +++++++++++++++-------
 1 file changed, 15 insertions(+), 7 deletions(-)

diff --git a/fs/ext4/resize.c b/fs/ext4/resize.c
index cb99b410c9fa..66ceabbd83ad 100644
--- a/fs/ext4/resize.c
+++ b/fs/ext4/resize.c
@@ -1110,6 +1110,16 @@ static int reserve_backup_gdb(handle_t *handle, struct inode *inode,
 	return err;
 }
 
+static inline void ext4_set_block_group_nr(struct super_block *sb, char *data,
+					   ext4_group_t group)
+{
+	struct ext4_super_block *es = (struct ext4_super_block *) data;
+
+	es->s_block_group_nr = cpu_to_le16(group);
+	if (ext4_has_metadata_csum(sb))
+		es->s_checksum = ext4_superblock_csum(sb, es);
+}
+
 /*
  * Update the backup copies of the ext4 metadata.  These don't need to be part
  * of the main resize transaction, because e2fsck will re-write them if there
@@ -1158,7 +1168,8 @@ static void update_backups(struct super_block *sb, sector_t blk_off, char *data,
 	while (group < sbi->s_groups_count) {
 		struct buffer_head *bh;
 		ext4_fsblk_t backup_block;
-		struct ext4_super_block *es;
+		int has_super = ext4_bg_has_super(sb, group);
+		ext4_fsblk_t first_block = ext4_group_first_block_no(sb, group);
 
 		/* Out of journal space, and can't get more - abort - so sad */
 		err = ext4_resize_ensure_credits_batch(handle, 1);
@@ -1168,8 +1179,7 @@ static void update_backups(struct super_block *sb, sector_t blk_off, char *data,
 		if (meta_bg == 0)
 			backup_block = ((ext4_fsblk_t)group) * bpg + blk_off;
 		else
-			backup_block = (ext4_group_first_block_no(sb, group) +
-					ext4_bg_has_super(sb, group));
+			backup_block = first_block + has_super;
 
 		bh = sb_getblk(sb, backup_block);
 		if (unlikely(!bh)) {
@@ -1187,10 +1197,8 @@ static void update_backups(struct super_block *sb, sector_t blk_off, char *data,
 		memcpy(bh->b_data, data, size);
 		if (rest)
 			memset(bh->b_data + size, 0, rest);
-		es = (struct ext4_super_block *) bh->b_data;
-		es->s_block_group_nr = cpu_to_le16(group);
-		if (ext4_has_metadata_csum(sb))
-			es->s_checksum = ext4_superblock_csum(sb, es);
+		if (has_super && (backup_block == first_block))
+			ext4_set_block_group_nr(sb, bh->b_data, group);
 		set_buffer_uptodate(bh);
 		unlock_buffer(bh);
 		err = ext4_handle_dirty_metadata(handle, NULL, bh);
-- 
2.31.1

[PATCH v3 3/3] ext4: fix corruption when online resizing a 1K bigalloc fs

[Baokun Li]

When a backup superblock is updated in update_backups(), the primary
superblock's offset in the group (that is, sbi->s_sbh->b_blocknr) is used
as the backup superblock's offset in its group. However, when the block
size is 1K and bigalloc is enabled, the two offsets are not equal. This
causes the backup group descriptors to be overwritten by the superblock
in update_backups(). Moreover, if meta_bg is enabled, the file system will
be corrupted because this feature uses backup group descriptors.

To solve this issue, we use a more accurate ext4_group_first_block_no() as
the offset of the backup superblock in its group.

Fixes: d77147ff443b ("ext4: add support for online resizing with bigalloc")
Signed-off-by: Baokun Li <libaokun1@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
---
V1->V2:
    Replace s_first_data_block with ext4_group_first_block_no() to avoid
    type warning.(Reported-by: kernel test robot <lkp@intel.com>)

 fs/ext4/resize.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/fs/ext4/resize.c b/fs/ext4/resize.c
index 66ceabbd83ad..b493233750ab 100644
--- a/fs/ext4/resize.c
+++ b/fs/ext4/resize.c
@@ -1604,8 +1604,8 @@ static int ext4_flex_group_add(struct super_block *sb,
 		int meta_bg = ext4_has_feature_meta_bg(sb);
 		sector_t old_gdb = 0;
 
-		update_backups(sb, sbi->s_sbh->b_blocknr, (char *)es,
-			       sizeof(struct ext4_super_block), 0);
+		update_backups(sb, ext4_group_first_block_no(sb, 0),
+			       (char *)es, sizeof(struct ext4_super_block), 0);
 		for (; gdb_num <= gdb_num_end; gdb_num++) {
 			struct buffer_head *gdb_bh;
 
@@ -1816,7 +1816,7 @@ static int ext4_group_extend_no_check(struct super_block *sb,
 		if (test_opt(sb, DEBUG))
 			printk(KERN_DEBUG "EXT4-fs: extended group to %llu "
 			       "blocks\n", ext4_blocks_count(es));
-		update_backups(sb, EXT4_SB(sb)->s_sbh->b_blocknr,
+		update_backups(sb, ext4_group_first_block_no(sb, 0),
 			       (char *)es, sizeof(struct ext4_super_block), 0);
 	}
 	return err;
-- 
2.31.1

Re: [PATCH v3 2/3] ext4: fix corrupt backup group descriptors after online resize

[Jan Kara]

On Thu 17-11-22 12:03:40, Baokun Li wrote:
> In commit 9a8c5b0d0615 ("ext4: update the backup superblock's at the end
> of the online resize"), it is assumed that update_backups() only updates
> backup superblocks, so each b_data is treated as a backupsuper block to
> update its s_block_group_nr and s_checksum. However, update_backups()
> also updates the backup group descriptors, which causes the backup group
> descriptors to be corrupted.
> 
> The above commit fixes the problem of invalid checksum of the backup
> superblock. The root cause of this problem is that the checksum of
> ext4_update_super() is not set correctly. This problem has been fixed
> in the previous patch ("ext4: fix bad checksum after online resize").
> 
> However, we do need to set block_group_nr for the backup superblock in
> update_backups(). When a block is in a group that contains a backup
> superblock, and the block is the first block in the group, the block is
> definitely a superblock. We add a helper function that includes setting
> s_block_group_nr and updating checksum, and then call it only when the
> above conditions are met to prevent the backup group descriptors from
> being incorrectly modified.
> 
> Fixes: 9a8c5b0d0615 ("ext4: update the backup superblock's at the end of the online resize")
> Signed-off-by: Baokun Li <libaokun1@huawei.com>

Looks good to me now. Feel free to add:

Reviewed-by: Jan Kara <jack@suse.cz>

								Honza

> ---
> V2->V3:
>     Modify the scheme and retain s_block_group_nr.
> 
>  fs/ext4/resize.c | 22 +++++++++++++++-------
>  1 file changed, 15 insertions(+), 7 deletions(-)
> 
> diff --git a/fs/ext4/resize.c b/fs/ext4/resize.c
> index cb99b410c9fa..66ceabbd83ad 100644
> --- a/fs/ext4/resize.c
> +++ b/fs/ext4/resize.c
> @@ -1110,6 +1110,16 @@ static int reserve_backup_gdb(handle_t *handle, struct inode *inode,
>  	return err;
>  }
>  
> +static inline void ext4_set_block_group_nr(struct super_block *sb, char *data,
> +					   ext4_group_t group)
> +{
> +	struct ext4_super_block *es = (struct ext4_super_block *) data;
> +
> +	es->s_block_group_nr = cpu_to_le16(group);
> +	if (ext4_has_metadata_csum(sb))
> +		es->s_checksum = ext4_superblock_csum(sb, es);
> +}
> +
>  /*
>   * Update the backup copies of the ext4 metadata.  These don't need to be part
>   * of the main resize transaction, because e2fsck will re-write them if there
> @@ -1158,7 +1168,8 @@ static void update_backups(struct super_block *sb, sector_t blk_off, char *data,
>  	while (group < sbi->s_groups_count) {
>  		struct buffer_head *bh;
>  		ext4_fsblk_t backup_block;
> -		struct ext4_super_block *es;
> +		int has_super = ext4_bg_has_super(sb, group);
> +		ext4_fsblk_t first_block = ext4_group_first_block_no(sb, group);
>  
>  		/* Out of journal space, and can't get more - abort - so sad */
>  		err = ext4_resize_ensure_credits_batch(handle, 1);
> @@ -1168,8 +1179,7 @@ static void update_backups(struct super_block *sb, sector_t blk_off, char *data,
>  		if (meta_bg == 0)
>  			backup_block = ((ext4_fsblk_t)group) * bpg + blk_off;
>  		else
> -			backup_block = (ext4_group_first_block_no(sb, group) +
> -					ext4_bg_has_super(sb, group));
> +			backup_block = first_block + has_super;
>  
>  		bh = sb_getblk(sb, backup_block);
>  		if (unlikely(!bh)) {
> @@ -1187,10 +1197,8 @@ static void update_backups(struct super_block *sb, sector_t blk_off, char *data,
>  		memcpy(bh->b_data, data, size);
>  		if (rest)
>  			memset(bh->b_data + size, 0, rest);
> -		es = (struct ext4_super_block *) bh->b_data;
> -		es->s_block_group_nr = cpu_to_le16(group);
> -		if (ext4_has_metadata_csum(sb))
> -			es->s_checksum = ext4_superblock_csum(sb, es);
> +		if (has_super && (backup_block == first_block))
> +			ext4_set_block_group_nr(sb, bh->b_data, group);
>  		set_buffer_uptodate(bh);
>  		unlock_buffer(bh);
>  		err = ext4_handle_dirty_metadata(handle, NULL, bh);
> -- 
> 2.31.1
> 
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR